Heuristic optimization of demand forecasting in a fuzzy environment

Abstract

forecasting the future demand of any kind of products of the business companies is a very important and critical activity for ail kinds of business organizations. For a retail chain store, demand forecasting is the major activity in their whole supply chain nctwork. Demand has to be accurately forecasted in order to fulfill thc customer requirements and in order to successfully run the business. An efficient and accurate demand forecasting system can playa major role in minimizing different kinds of costs and in increasing customer service. That mcans overall quality of the organization can bc increased. There are hundrcds of different techniques have been invented so far for efficient demand forecasting. Some arc qualitative and some are quantitative methods. There are also some mcthods which ore combination of both. Customer comes to chain retail store in order to buy products. Making products a~ailable for the customer for buying is the objective of the management of the company. Hundreds of I'arielles products are available in a chain retail store, There is various demand influencing factors for different kinds of products, It is very much difficult to include thc quantitative effects of the influcncing factors in the demand of any kind of item by applylllg the existing forecasting algoritlullS. Though, the current algorithms use both quantitative and qualitative mcthods. in their forecasting techniques, there some limltations of the existing algonthms, Thcre are such influencing factors in the dcmand of the items available in a chain retail store whose effccts can not be quantified by the existing algorithms. Neural network is a ,-ery promising tool in the field of forecasting. It is a data driven method, It ciln identify paltem in the past data and base on that pattern it can predict or forecast the futliTe data, Though, il is a quantitative method, judgmentat decisions can be applled through this method. Forecasting using artificial neural network technique is the most advanced procedure in any killd of forecasting field, Applymg altifieial neural net,,'ork algoritlun III retail store demand forecasting is a very challenging lask. In this sllldy, the artificial neural network algorithm has been applied for forecasting future demand of a fast moving ilem in a chain retail store, Previous years demand data has been used in developing the algorithm. The demand pattern of the selected item has been studied' initially. Network' archi'tccture - haS -been- creaiC{j-'hy'- ~s-i;;g -tl{e observations of that study. The Tesult has found to be very mueh encouraging. At the begilUling of this research, It has been reviewed that in the retail sector the error of the curreut forecasting algorithms is the range of 20% to 25%. The algoritlun that has been developed in this study tlle error is about 8% to 10%. The reduction of forecasting error will definitely contribute in tile development of the chain retail stores and achieving higher profit and customer satisfaction level.